When Providers Use AI to Bill Better, Payment Integrity Has to Get Smarter

Better documentation from providers can look a lot like upcoding. The latest healthcare AI developments are creating new challenges for payment integrity programs — and inspiring new solutions.

As healthcare costs accelerate, providers are using increasingly sophisticated AI tools for a range of tasks: ambient documentation, coding, revenue optimization, and more. From the provider’s perspective, this is a good thing: better documentation captures care that was previously undercoded, enabling them to be paid more appropriately.

For payers, however, these rapid shifts in provider AI billing and coding behavior raise an important question: is the claim more accurate, or simply better optimized for reimbursement?

Both perspectives are legitimate. Theoretically, these AI documentation and billing tools are designed to make clinicians more efficient and capture more data, resulting in more specificity, more accurate coding, and appropriate acuity. But they can also be systematically applied to identify areas where reimbursement can be maximized. Emerging research also raises questions about accuracy, missing context, and how the tools are impacting consumer healthcare costs.

Kristina Johnson Rose, Director of Payment Integrity at Premera Blue Cross, noticed the shift right away: “It just seemed to happen overnight.”  She initially suspected providers were upcoding. In some cases, she says, they were just finally billing correctly. But she determined others were intentionally using billing practices to pursue additional revenue.

That’s the dilemma. AI-generated documentation is changing provider billing behaviors. Those behaviors are not always intentional or malicious. But without visibility into the full context behind these new billing trends, payers can’t always tell the difference.

Figuring Out the “Why”

The reasons behind changes in coding intensity trends tend to be more complicated than simply “AI upcoding.” To try to evaluate whether billing is more accurate or more aggressive when coding intensity increases, payers can use a three-part framework for evaluation:

1. If the care was previously being delivered but not consistently documented or coded, the reimbursement increase is likely due to more complete and accurate documentation and therefore legitimate.
2. If the provider’s patient mix, services, or treatment patterns genuinely changed, that is likely due to changing clinical complexity, another legitimate reason for reimbursement increases.
3. If technology is identifying reimbursement opportunities that seem to exceed what the underlying care supports, revenue optimization tactics may be at play.

All three situations can show up as an increase in coding intensity in claims data, making static threshold-based approaches to identifying potential upcoding less effective.

This is happening across coding scenarios. “We’re seeing a lot in the professional claims now,” says Dr. Priscilla Alfaro, a strategic PI advisor. “People talk about the DRG coding, but we’re also seeing that in our APC coding.”

Why More Audits Aren’t the Answer

When documentation and coding behavior changes before submission, retrospective recovery is just a band-aid covering the real problem. The better solution is to move PI upstream.

The shift to prepay is already well underway, but “we have additional opportunities to move things even ahead of prepay,” says Dr. Alfaro, including a growing focus on pre-submission as a method to identify preventive savings before a claim is even submitted.

“I’m most looking forward to the continued shift to prepay and the collaboration between the providers and payers, but also what we are going to see with the move to a focus on pre-submission.”

— Susan Luden, Director of Financial Investigations & Provider Review, Highmark Inc.

But the goal isn’t to create an AI arms race in which providers continue to optimize technology and are met with increasingly aggressive detection tactics by payers. Instead, payers and providers should work together to ensure accurate billing.

Payers could even seize this trend as a teaching moment, moving the PI framework from “find, deny, recover” toward “understand, validate, educate, prevent.”

“This is your opportunity to teach your provider how to bill.”

— Gladys Oswell, Founder and CEO, Oswell Healthcare Advisory

Continuous Monitoring > Occasional Audits

Because coding practices, technology, reimbursement strategies, and definitions of low-value work continually change, building an algorithm around today’s coding behavior is not an effective solution for the long term. Instead, payers should institute continuous monitoring that looks for:

• Emerging shifts in coding behavior
• Provider outliers
• Patterns that can’t be explained by clinical or population changes
• Isolated vs. systemic changes
• Changes requiring clinical validation

This ongoing visibility into claims behavior then becomes a competitive advantage over payers that continue to rely on a library of edits.

“We’ve hit a point with providers where it behooves us to be collaborative and work together to solve these issues, as opposed to taking sides.”

— Dr. Priscilla Alfaro, Payment Integrity Leader

Evolving AI Requires More Context

As providers continue to integrate AI, the less useful it is for payers to view claims in isolation. Determining payment accuracy increasingly requires connecting the dots between the claim, the provider, clinical information, historical behavior, the contract, coding pattern, and the payment outcome.

“You have 300, 400 algorithms, and you just keep coming up with these ideations. The next step is using AI to get through these items and let that technology build out what you’re looking for.”

— Novelette Wallace, Head of Payment and Program Integrity, Johns Hopkins Health Plans

This evolution highlights why disconnected point solutions aren’t the answer. They do not provide the integrated data and operational visibility this complex and dynamic environment requires. By connecting claims, audit data, workflows, vendors, and results, Pareo® provides the comprehensive context in which payers can surface, evaluate, and act on changes in coding behavior at scale.

A Vision for the Future of PI

Instead of focusing only on continually refining methods for identifying inaccurate payments, payers should work with providers to create an environment where accurate payments are the default. This new paradigm looks like:

• Providers documenting and billing correctly
• Payers recognizing legitimate clinical complexity
• Identifying inappropriate reimbursement before payment
• Contracts that reflect changing realities
• Technology that handles repetitive tasks
• People focusing on exceptions that require human judgment

As Kristina Johnson Rose puts it, “the best PI is no PI.”

Integrating AI tools alone won’t lead to this outcome. It requires health plans to adopt a more proactive and collaborative approach that incorporates visibility, context, and continuous monitoring. The ones that do so will be the plans that can see what’s changing, understand why, and take action before payment inaccuracy becomes payment waste.

Talk to ClarisHealth about how the Pareo® payment integrity technology platform is helping health plans gain greater visibility and control across the payment integrity lifecycle.